I recently provided some thoughts to journalist Hazel Davies, who was writing about AI Start Ups and the music business for The Guardian newspaper. Here is the link to the full article, Robot rhythms: the startups using AI to shake up the music business, but here I’m pasting below the full text I provided to Hazel.
This text came from some thinking I’ve been engaged with recently as I’m in the early stages of developing a research funding bid around AI and Music with some of my BCU colleagues. More news on that soon, I hope, but in the meantime you may find this of interest if you are looking at the role of AI in popular music.
AI and Music is certainly a growth area and likely to continue to be so. There are a number of start-ups who have attracted investment over the last couple of years - companies such as Amper and AI Music - and there is also a lot of activity within many of the larger tech and entertainment firms, notably Sony, Google, Spotify, IBM and Facebook. This perhaps should not be surprising - AI systems rely on large amounts of data to build and train models, and on powerful computational, connected systems to process data and deploy the results. Since the turn of the century music consumption has increasingly shifted to systems capable of generating and analysing data on both user activity and musical works, and there has also been an exponential growth in the scale and power of computational processing systems. In that sense, AI and music together does have a certain air of inevitability about it. However, we are still in the very early stages of that relationship and it remains to be seen if and how it will develop.
The business models of the emerging AI Music companies have tended to focus mainly on B2B propositions - i.e. services that can either create music - for use in games, video, and other media - at a much cheaper rate than existing synchronisation channels, or else on services positioned as tools for producers and songwriters based on the promise of augmenting the creative process. Clearly the first scenario has potential implications for songwriters, producers and rights holders, who derive some or all of their income from synchronisation. The second scenario could also be of concern to songwriters and producers, because several of the vendors of the AI systems mentioned above make the claim that ‘anyone’ will soon be able to create a highly sophisticated musical work. This is where things get interesting for the consumer, because many of the companies are also hinting at the potential for more mass-market applications - for example, the ability for social media users to also create ‘royalty free’ music for their day-to-day video and photo posts, or else for ‘personalised’ playlists of music that will react to moods, scenarios, or locations, in real-time. If that latter scenario comes to pass then it’s likely that the more interesting/better performing start-ups focussing the the B2C end will be quickly acquired by the larger tech firms, which is presumably what their VC investors are banking on.
For the B2B end the bigger questions concern ideas of ownership and creativity. The recorded music business is fundamentally based on the ownership and exploitation of rights that hinge on the recognition of individual(s) as original creators. What happens, then, if a songwriter/producer uses an AI-based system to create a song that becomes a global hit on the scale of something like ‘Get Lucky’? Who is the songwriter and who is the performer in that instance, and who - crucially - owns which rights? Creators using such systems should therefore closely examine the terms and conditions they are binding themselves and their work to, as indeed should the rights holders of existing works - i.e. the music created by humans, pre-AI - because it is their music that is being processed and analysed in order to build AI systems. Would the owners of an AI system trained on the catalogue of The Beatles, for instance, owe any sort of creative/financial debt to Lennon and McCartney? For the rest of us - the ordinary consumers using systems to produce material for use on social media, or else listening to AI-created playlists in the gym - what are the longer-term implications on our already fragile data privacy, for our intellectual property rights, or - more broadly - for the culture of popular music of which we are all a part? When these new, cool apps begin to appear on the market we should perhaps ask ourselves those questions before diving in.